CN105426895A - Prominence detection method based on Markov model - Google Patents
Prominence detection method based on Markov model Download PDFInfo
- Publication number
- CN105426895A CN105426895A CN201510760313.1A CN201510760313A CN105426895A CN 105426895 A CN105426895 A CN 105426895A CN 201510760313 A CN201510760313 A CN 201510760313A CN 105426895 A CN105426895 A CN 105426895A
- Authority
- CN
- China
- Prior art keywords
- obtains
- image
- rectangular area
- significant characteristics
- function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Probability & Statistics with Applications (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a prominence detection method based on the Markov model. The prominence detection method comprises the following steps: first, calculate prominence characteristic diagrams via three different prominences of multi-scale comparison, central-peripheral histogram and color spatial distribution; then, studying and calculating the weight of each prominence characteristic diagram via the Markov model, and acquiring a model parameter estimation optimal solution according to the maximum likelihood estimation method; and finally, detecting an image to be tested via the Markov model. The method provided by the invention can be used for detecting a prominent object more exactly, and is high in detection result resolution, accurate in object border definition, and low in calculation complexity.
Description
Technical field
The present invention relates to a kind of conspicuousness detection method based on Markov model, belong to image procossing and image object detection technique field.
Background technology
Vision is the most important sensory perceptions of the mankind, human brain receptible external information more than 90% come from the visually-perceptible of human eye.The major function of vision is exactly explain the surrounding environment of people's life, and with its generation information interaction, developing rapidly of infotech impels various image information day by day to expand, and people's computer system of having to processes and analyze these mass datas.But it should be noted that: on the one hand, gathering way of view data is wanted soon more than the raising speed of computer process ability; On the other hand, the content that people are concerned about is a part very little in whole data acquisition usually.For this reason, all view data of overall treatment of making no exception are unpractical, are also unnecessary.How find from whole data centralization as soon as possible and extract that the part relevant to task is important, information that is useful and that merit attention, i.e. vision significance test problems is exactly machine vision and the medium-term and long-term important problem faced of information processing research always.Be that the saliency region detection technology of representative becomes one of important technology approach improving mass data screening real-time and precision of analysis with vision attention.It is an important content in image procossing that conspicuousness detects, and has a wide range of applications, as the Iamge Segmentation based on conspicuousness, and image retrieval, image automatic cutting and the compression of image screen etc.
The essence that conspicuousness detects is a kind of visual attention model, and this model is the model set up according to vision noticing mechanism, and it can distribute limited Information procession resource, makes perception possess selective power.Utilize vision noticing mechanism to obtain the signal portion the most easily aroused attention in image, and represent its significance with a width gray level image.Psychology of vision research finds, human visual attention can be divided into two types: bottom-up data driven mode and top-down task driving teaching method.Bottom-up is at the visual processes initial stage, and be not subject to the impact of experience and current task, the mankind exist particular concern region and salient region to scene.Top-down is the visual processes later stage, and the experience of the mankind according to self and the target of task choosing concern, be familiar with target.
Early stage its major defect of saliency algorithm is that resolution is low, and object boundary definition is poor, and computation complexity is high.
Summary of the invention
Goal of the invention: for problems of the prior art, the invention provides the conspicuousness detection method based on Markov model that a kind of resolution is high, computation complexity is high.
Technical scheme: a kind of conspicuousness detection method based on Markov model, comprises the steps:
Step 10: acquisition of image data;
Step 20, carries out significant characteristics extraction to the image that step 10 obtains with three kinds of diverse ways, obtains the significant characteristics figure that the significant characteristics function different from three kinds is corresponding;
Step 30: adopt the machine learning method of Markov model to train the image gathered in step 10, and obtain the optimal weights of each significant characteristics figure obtained in step 20;
Step 40: three kinds that obtain step 20 different significant characteristics function partition function Z are normalized, three the normalization significant characteristics functions obtained;
Step 50: set up Markov model, combines three normalization significant characteristics function Markov models that step 40 obtains;
Step 60: try to achieve optimum solution to the combination that step 50 obtains by maximum-likelihood criterion, obtains optimized linear combination;
Step 70: conspicuousness pixel step 60 calculated a minimum rectangle circle goes out, and wherein minimum rectangle frame at least frame goes out the conspicuousness pixel of more than 95%, obtains final result.
Further, in described step 20, described three kinds of methods of carrying out feature extraction are respectively: multiple dimensioned pairing comparision, central authorities histogram method and Color-spatial distribution method around;
Wherein, described multiple dimensioned pairing comparision, comprises the following steps:
Step 211, to the image collected in step 10 based on down sample after Gaussian Blur, obtains the gaussian pyramid image of six layers of different resolution;
Step 212, in this pyramid diagram picture of six floor heights step 211 obtained, every layer of contrast linear combination obtains multiple dimensioned contrast characteristic's function significant characteristics figure corresponding with it;
Described central authorities histogram method around, comprises the following steps:
Step 221: mark the obvious object in the image that step 10 obtains with the rectangular area R of multiple different Aspect Ratio, rectangular area R around the area equation of the multiple correspondence of surrounding structure of multiple rectangular area R
s;
Step 222: calculate on the image that step 221 obtains each remarkable rectangular area R centered by pixel x and around rectangular area R
sχ between RGB color histogram
2distance;
Step 223: the rectangular area R of more each different Aspect Ratio and surrounding rectangular area R
sχ between RGB color histogram
2distance, selects χ
2apart from the rectangular area R that maximum rectangular area R is optimum
*(x);
Step 224: the center-periphery histogram feature function centered by neighbor x ' is defined as all optimum rectangular area R around centered by neighbor x ' in step 221
s *gauss's weighting χ of (x ')
2distance sum;
Described Color-spatial distribution method, comprises the following steps:
Step 231: all colours in the image obtain step 10 represents with gauss hybrid models;
Step 232: utilize the parameter of model in step 231 to calculate the probability that each pixel is assigned to a kind of color component;
Step 233: corresponding level variance and vertical variance are calculated to each color component in step 232, obtains the space variance of tie element;
Step 234: Color-spatial distribution fundamental function is defined as the space variance central authorities weighting sum that step 233 obtains.
Further, in step 50, described Markov model combination significant characteristics process is as follows:
Step 501, to three normalization significant characteristics that step 40 obtains, calculates unitary potential function F respectively
k(a
x, I), F
k(a
x, I) and represent a kth notable feature;
Step 502, to three normalization significant characteristics that step 40 obtains, calculates binary potential function S (a respectively
x, a
x', I) and match feature, wherein, binary potential function S (a
x, a
x', I) and represent the value of penalty term neighbor being labeled as to different value, a
xrepresent the conspicuousness of x pixel, a
x'represent the conspicuousness of x neighbor pixel;
Step 503, the notable feature obtained in the optimal weights of each significant characteristics figure that integrating step 30 obtains and step 501,502 and pairing feature, according to formula
Carry out linear combination, wherein, A is for gathering label state set in image I, and Z is partition function, λ
krepresent the weight of a kth notable feature figure, K is the sum of notable feature figure, F
k(a
x, I) and be the potential function of unitary variant, F
k(a
x, I) and represent kth notable feature figure, a S (a
x, a
x', I) be the mutual potential function of bivariate, S (a
x, a
x', I) represent neighbor x, the interaction relationship between x '.
Principle of work: conspicuousness detects and regards an image labeling problem as by the present invention, uses multiple dimensioned contrast, the significance that central authorities-around histogram is different with Color-spatial distribution these three kinds calculates notable feature figure.Calculated the weight of the significance of each notable feature figure by Markov model study, adopt maximum Likelihood to obtain model parameter estimation and obtain optimum solution.Markov model is finally utilized to detect test pattern.
Beneficial effect: compared with prior art, method provided by the invention more accurately can detect well-marked target, detects the result resolution obtained high, and precisely, method computation complexity is low in object boundary definition.
Accompanying drawing explanation
Fig. 1 is process flow diagram of the present invention;
Fig. 2 is that in the present invention, significant characteristics extracts process flow diagram;
Fig. 3 is that the Experimental comparison of the present invention and prior art schemes.
Embodiment
Below in conjunction with specific embodiment, illustrate the present invention further, these embodiments should be understood only be not used in for illustration of the present invention and limit the scope of the invention, after having read the present invention, the amendment of those skilled in the art to the various equivalent form of value of the present invention has all fallen within the application's claims limited range.
As shown in Figure 1, based on the conspicuousness detection method of Markov model, comprise the steps:
Step 10: acquisition of image data; It is I that definition gathers image.
Step 20, carries out significant characteristics extraction to the image I that step 10 obtains with three kinds of diverse ways, obtains the significant characteristics figure that the significant characteristics function different from three kinds is corresponding;
Around histogram method and Color-spatial distribution method carry out feature extraction to adopt multiple dimensioned pairing comparision, central authorities in the present embodiment.
1, multiple dimensioned pairing comparision
In conspicuousness detects, contrast method is the most often used on local feature.When not knowing remarkable object size, the conspicuousness that we adopt multi-scale method to carry out regional area respectively detects.Mainly comprise the following steps:
Step 211, to the image collected in step 10 based on down sample after Gaussian Blur, obtains the image of different resolution; The new figure image width simultaneously at every turn obtained and height are 1/2 of original images, and a series of images obtained is called gaussian pyramid.
Step 212, in this pyramid diagram picture of six floor heights step 211 obtained, every layer of contrast linear combination obtains multiple dimensioned contrast characteristic's function
the significant characteristics figure corresponding with it; Wherein, I
lthe l tomographic image in pyramid, I
lx () represents the state value of pixel x on l tomographic image, i.e. I
lwhen () equals 0 x, represent that on l tomographic image, pixel x is conspicuousness pixel, I
lwhen () equals 1 x, represent that pixel x is non-limiting pixel on l tomographic image; I
l(x ') represents the state value of the neighbor x ' of pixel x on l tomographic image, and the number of plies that pyramid has altogether is the window that L=6, N (x) represent 9x9.
2, central authorities' histogram method around
On given RGB color space basis, the pixel adding up often kind of color component accounts for the ratio of the total pixel of image, thus obtains ratio distribution and the histogram of image shades of colour component.Suppose that an obvious object is gone out by rectangular area R frame, around it, we construct a rectangular area R with homalographic
s.Herein by calculating remarkable rectangular area R centered by pixel x and rectangular area R around it
s, their χ between RGB color histogram
2distance represents conspicuousness.Because the target size of conspicuousness object is different, we select the rectangular area of different length breadth ratios to test.Mainly comprise the following steps:
Step 221: mark the obvious object in the image that step 10 obtains with the rectangular area R of five groups of different Aspect Ratios, rectangular area R around the area equation of the multiple correspondence of surrounding structure of multiple rectangular area R
s; Wherein, five groups of different Aspect Ratios are { 0.5,0.75,1.0,1.5,2};
Step 222: calculate on the image that step 221 obtains each remarkable rectangular area R centered by pixel x and around rectangular area R
sχ between RGB color histogram
2distance;
Step 223: the rectangular area R of more each different Aspect Ratio and the surrounding rectangular area R of rectangular area R homalographic
sχ between RGB color histogram
2distance, selects χ
2apart from the rectangular area R that maximum rectangular area R is optimum
*(x);
Step 224: the center-periphery histogram feature function centered by neighbor x ' is defined as all optimum rectangular area R around centered by neighbor x ' in step 221
s *gauss's weighting χ of (x ')
2distance sum; Center-periphery histogram feature function is:
Wherein weight w
xx '=exp (-0.5 σ
x ' -2|| x-x ' ||
2), represent by Gauss's damping capacity, σ
x ' 2represent the covariance of pixel x', R
*(x ') represents optimum rectangular area centered by neighbor x '.
3, Color-spatial distribution method
Center-periphery histogram only describes the characteristic area of a local, but the Color-spatial distribution of the overall situation also can describe conspicuousness object in piece image, therefore present invention incorporates the feature of local and the overall situation to carry out conspicuousness detection.
Step 231: to step: all colours in 10 images obtained represents with gauss hybrid models; The simplest method of space distribution describing a color is exactly the space variance calculating color, adopts gauss hybrid models (GMM) here.All colours in image GMM represents, this model has three parameters
wherein w
crepresent the weight of c color, u
crepresent the color average of c color, Σ
crepresent the covariance matrix of c color component, C represents the sum of color.
Step 232: utilize the parameter of model in step 231 to calculate the probability that each pixel is assigned to a kind of color component:
Wherein, N (I
x| u
c, Σ
c) be a Gaussian distribution, for judging whether pixel x belongs to c color component.
Step 233: corresponding level variance V is calculated to each color component in step 232
h(c) and vertical variance V
vc (), obtains the space variance of tie element | X|
c=V
v(c)+V
h(c);
Step 234: Color-spatial distribution fundamental function is defined as space variance central authorities weighting that step 433 obtains
Sum
Step 30: set up Markov model, adopts the machine learning method of Markov model to train the image gathered in step 10, and obtains the optimal weights of each significant characteristics figure obtained in step 20; The optimal weights λ obtained in the present embodiment
*={ 0.25,0.48,0.27}
Step 40: three kinds that obtain step 20 different significant characteristics function partition function Z are normalized;
Three kinds of different significant characteristics define a notable feature function f respectively
x(x, I), is then normalized f
x(x, I) ∈ [0,1].Notable feature is defined as follows:
F
k(a
x,I)=f
x(x,I)a
x=0
Wherein, a
x=0 represents that x pixel is conspicuousness, a
x=1 represents that x pixel is non-limiting.
The mutual potential function of bivariate is expressed as follows:
S(a
x,a
x',I)=|a
x-a
x′|·exp(-βd
x,x′)
Wherein, d
x, x '=|| I
x-I
x '|| be that two norms of colour-difference represent, β=(2<||I
x-I
x '||
2>)
-1represent color contrast weight parameter.
Step 50: three normalization significant characteristics function Markov models that step 40 obtains are combined;
Step 60: try to achieve optimum solution to the combination that step 50 obtains by maximum-likelihood criterion, obtains optimized linear combination;
In order to obtain optimum characteristic line combination, to N width training image
maximal possibility estimation is used to train, wherein A
nrepresent the n-th width training image I
nremarkable figure, n represent the n-th width training image, after getting log to it, expression formula is convex function, has optimum solution as follows:
After obtaining optimum parameter, derived by Markov model, try to achieve final optimum solution:
y
*=argmaxP(A|I)
Step 70: step 60 is calculated a series of state value, wherein, 0 represents conspicuousness pixel, 1 represents non-limiting pixel, go out conspicuousness pixel by a minimum rectangle circle, wherein minimum rectangle frame at least frame goes out the conspicuousness pixel of more than 95%, obtains final result.
By multiple dimensioned pairing comparision obtained above, central authorities' feature that around histogram method and Color-spatial distribution method obtain carries out linear combination, according to confession by training the optimal weights obtained
Obtain the flag sequence that probability of occurrence is maximum, flag sequence gray level image is showed thus final remarkable figure can be obtained.
By above-mentioned embodiment, visible tool of the present invention has the following advantages: in order to weigh the present invention put forward the validity of algorithm, we can use assessment indicator: recall rate (Recall), and accurate rate (Precision) and F-value (F-measure) compare.As shown in Figure 3, the multiple dimensioned contrast in local will be adopted separately respectively, center-periphery histogram and spatial color distribution feature are through testing the assessment indicator obtained, and finally the assessment indicator that three fundamental functions are obtained by best initial weights linear combination compares by our method.The experimental result adopting separately the multiple dimensioned contrast in local in Fig. 3 is 1, adopts the histogrammic experimental result of center-periphery to be 2 separately, and the experimental result adopting spatial color to divide separately is 3, is 4 by three fundamental functions by the test findings of best initial weights linear combination.From Fig. 3 we can find out adopt the result that obtains of multiple dimensioned contrast characteristic be accurate rate very high while recall rate also very low, because obvious object interior zone is same item, so the contrast of inside is not high thus result in lower recall rate.Center-periphery histogram obtains a good F-value in this several method, although this feature interpretation comprises the pixel of some mistakes in background area, this local feature can detect complete obvious object well.Spatial color distribution has lower accurate rate and the highest recall rate.In the conspicuousness of our research detects, the index of recall rate does not have accurate rate important.Obtain in this method by three kinds of characteristic line combinations, accurate rate, the index of recall rate and F-value is all higher.Present invention incorporates local and global characteristics, by Markov model learning training parameter, three kinds of different characteristic line combinations are obtained optimum result.Show by experiment, this method is with former Measures compare, and this method can detect obvious object more accurately.
Claims (3)
1., based on a conspicuousness detection method for Markov model, it is characterized in that, comprise the steps:
Step 10: acquisition of image data;
Step 20, carries out significant characteristics extraction to the image that step 10 obtains with three kinds of diverse ways, obtains the significant characteristics figure that the significant characteristics function different from three kinds is corresponding;
Step 30: adopt the machine learning method of Markov model to train the image gathered in step 10, and obtain the optimal weights of each significant characteristics figure obtained in step 20;
Step 40: three kinds that obtain step 20 different significant characteristics function partition function Z are normalized, three the normalization significant characteristics functions obtained;
Step 50: set up Markov model, combines three normalization significant characteristics function Markov models that step 40 obtains;
Step 60: try to achieve optimum solution to the combination that step 50 obtains by maximum-likelihood criterion, obtains optimized linear combination;
Step 70: conspicuousness pixel step 60 calculated a minimum rectangle circle goes out, and wherein minimum rectangle frame at least frame goes out the conspicuousness pixel of more than 95%, obtains final result.
2. as claimed in claim 1 based on the conspicuousness detection method of Markov model, it is characterized in that, in described step 20, described three kinds of methods of carrying out feature extraction are respectively: multiple dimensioned pairing comparision, central authorities histogram method and Color-spatial distribution method around;
Wherein, described multiple dimensioned pairing comparision, comprises the following steps:
Step 211, to the image collected in step 10 based on down sample after Gaussian Blur, obtains the gaussian pyramid image of six layers of different resolution;
Step 212, in this pyramid diagram picture of six floor heights step 211 obtained, every layer of contrast linear combination obtains multiple dimensioned contrast characteristic's function significant characteristics figure corresponding with it;
Described central authorities histogram method around, comprises the following steps:
Step 221: mark the obvious object in the image that step 10 obtains with the rectangular area R of multiple different Aspect Ratio, rectangular area R around the area equation of the multiple correspondence of surrounding structure of multiple rectangular area R
s;
Step 222: calculate on the image that step 221 obtains each remarkable rectangular area R centered by pixel x and around rectangular area R
sχ between RGB color histogram
2distance;
Step 223: the rectangular area R of more each different Aspect Ratio and surrounding rectangular area R
sχ between RGB color histogram
2distance, selects χ
2apart from the rectangular area R that maximum rectangular area R is optimum
*(x);
Step 224: the center-periphery histogram feature function centered by neighbor x ' is defined as all optimum rectangular area R around centered by neighbor x ' in step 221
s *gauss's weighting χ of (x ')
2distance sum;
Described Color-spatial distribution method, comprises the following steps:
Step 231: all colours in the image obtain step 10 represents with gauss hybrid models;
Step 232: utilize the parameter of model in step 231 to calculate the probability that each pixel is assigned to a kind of color component;
Step 233: corresponding level variance and vertical variance are calculated to each color component in step 232, obtains the space variance of tie element;
Step 234: Color-spatial distribution fundamental function is defined as the space variance central authorities weighting sum that step 233 obtains.
3. as claimed in claim 1 based on the conspicuousness detection method of Markov model, it is characterized in that, in step 50, described Markov model combination significant characteristics process is as follows:
Step 501, to three normalization significant characteristics that step 40 obtains, calculates unitary potential function F respectively
k(a
x, I), F
k(a
x, I) and represent a kth notable feature;
Step 502, to three normalization significant characteristics that step 40 obtains, calculates binary potential function S (a respectively
x, a
x', I) and match feature, wherein, binary potential function S (a
x, a
x', I) and represent the value of penalty term neighbor being labeled as to different value, a
xrepresent the conspicuousness of x pixel, a
x'represent the conspicuousness of x neighbor pixel;
Step 503, the notable feature obtained in the optimal weights of each significant characteristics figure that integrating step 30 obtains and step 501,502 and pairing feature, according to formula
Carry out linear combination, wherein, A is for gathering label state set in image I, and Z is partition function, λ
krepresent the weight of a kth notable feature figure, K is the sum of notable feature figure, F
k(a
x, I) and be the potential function of unitary variant, F
k(a
x, I) and represent kth notable feature figure, a S (a
x, a
x', I) be the mutual potential function of bivariate, S (a
x, a
x', I) represent neighbor x, the interaction relationship between x '.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510760313.1A CN105426895A (en) | 2015-11-10 | 2015-11-10 | Prominence detection method based on Markov model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510760313.1A CN105426895A (en) | 2015-11-10 | 2015-11-10 | Prominence detection method based on Markov model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105426895A true CN105426895A (en) | 2016-03-23 |
Family
ID=55505094
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510760313.1A Pending CN105426895A (en) | 2015-11-10 | 2015-11-10 | Prominence detection method based on Markov model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105426895A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106127210A (en) * | 2016-06-17 | 2016-11-16 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A kind of significance detection method based on multiple features |
CN106651937A (en) * | 2016-10-19 | 2017-05-10 | 成都电科智达科技有限公司 | Method for detecting small unmanned aerial vehicle target based on super-pixels and scene prediction |
CN106780430A (en) * | 2016-11-17 | 2017-05-31 | 大连理工大学 | A kind of image significance detection method based on surroundedness and Markov model |
CN107145824A (en) * | 2017-03-29 | 2017-09-08 | 纵目科技(上海)股份有限公司 | A kind of lane line dividing method and system, car-mounted terminal based on significance analysis |
CN107679516A (en) * | 2017-10-25 | 2018-02-09 | 中国科学院合肥物质科学研究院 | Lower extremity movement recognition methods based on multiple dimensioned Gauss Markov random field model |
CN108921833A (en) * | 2018-06-26 | 2018-11-30 | 中国科学院合肥物质科学研究院 | A kind of the markov conspicuousness object detection method and device of two-way absorption |
CN110097569A (en) * | 2019-04-04 | 2019-08-06 | 北京航空航天大学 | Oil tank object detection method based on color Markov Chain conspicuousness model |
CN110751157A (en) * | 2019-10-18 | 2020-02-04 | 厦门美图之家科技有限公司 | Image saliency segmentation and image saliency model training method and device |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120328161A1 (en) * | 2011-06-22 | 2012-12-27 | Palenychka Roman | Method and multi-scale attention system for spatiotemporal change determination and object detection |
CN104077609A (en) * | 2014-06-27 | 2014-10-01 | 河海大学 | Saliency detection method based on conditional random field |
CN104240256A (en) * | 2014-09-25 | 2014-12-24 | 西安电子科技大学 | Image salient detecting method based on layering sparse modeling |
CN104424642A (en) * | 2013-09-09 | 2015-03-18 | 华为软件技术有限公司 | Detection method and detection system for video salient regions |
-
2015
- 2015-11-10 CN CN201510760313.1A patent/CN105426895A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120328161A1 (en) * | 2011-06-22 | 2012-12-27 | Palenychka Roman | Method and multi-scale attention system for spatiotemporal change determination and object detection |
CN104424642A (en) * | 2013-09-09 | 2015-03-18 | 华为软件技术有限公司 | Detection method and detection system for video salient regions |
CN104077609A (en) * | 2014-06-27 | 2014-10-01 | 河海大学 | Saliency detection method based on conditional random field |
CN104240256A (en) * | 2014-09-25 | 2014-12-24 | 西安电子科技大学 | Image salient detecting method based on layering sparse modeling |
Non-Patent Citations (1)
Title |
---|
张微: "基于概率图模型的图像分类研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106127210A (en) * | 2016-06-17 | 2016-11-16 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A kind of significance detection method based on multiple features |
CN106651937A (en) * | 2016-10-19 | 2017-05-10 | 成都电科智达科技有限公司 | Method for detecting small unmanned aerial vehicle target based on super-pixels and scene prediction |
CN106651937B (en) * | 2016-10-19 | 2019-10-18 | 成都电科智达科技有限公司 | A kind of small drone object detection method based on super-pixel and scene prediction |
CN106780430B (en) * | 2016-11-17 | 2019-08-09 | 大连理工大学 | A kind of image significance detection method based on surroundedness and Markov model |
CN106780430A (en) * | 2016-11-17 | 2017-05-31 | 大连理工大学 | A kind of image significance detection method based on surroundedness and Markov model |
CN107145824A (en) * | 2017-03-29 | 2017-09-08 | 纵目科技(上海)股份有限公司 | A kind of lane line dividing method and system, car-mounted terminal based on significance analysis |
CN107679516B (en) * | 2017-10-25 | 2019-07-19 | 中国科学院合肥物质科学研究院 | Lower extremity movement recognition methods based on multiple dimensioned Gauss-Markov random field models |
CN107679516A (en) * | 2017-10-25 | 2018-02-09 | 中国科学院合肥物质科学研究院 | Lower extremity movement recognition methods based on multiple dimensioned Gauss Markov random field model |
CN108921833A (en) * | 2018-06-26 | 2018-11-30 | 中国科学院合肥物质科学研究院 | A kind of the markov conspicuousness object detection method and device of two-way absorption |
CN108921833B (en) * | 2018-06-26 | 2022-03-25 | 中国科学院合肥物质科学研究院 | Two-way absorption Markov significance target detection method and device |
CN110097569A (en) * | 2019-04-04 | 2019-08-06 | 北京航空航天大学 | Oil tank object detection method based on color Markov Chain conspicuousness model |
CN110751157A (en) * | 2019-10-18 | 2020-02-04 | 厦门美图之家科技有限公司 | Image saliency segmentation and image saliency model training method and device |
CN110751157B (en) * | 2019-10-18 | 2022-06-24 | 厦门美图之家科技有限公司 | Image significance segmentation and image significance model training method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105426895A (en) | Prominence detection method based on Markov model | |
CN104077609A (en) | Saliency detection method based on conditional random field | |
CN108830285B (en) | Target detection method for reinforcement learning based on fast-RCNN | |
CN104392463B (en) | Image salient region detection method based on joint sparse multi-scale fusion | |
CN103996018B (en) | Face identification method based on 4DLBP | |
CN105335716B (en) | A kind of pedestrian detection method extracting union feature based on improvement UDN | |
CN105160310A (en) | 3D (three-dimensional) convolutional neural network based human body behavior recognition method | |
CN109118479A (en) | Defects of insulator identification positioning device and method based on capsule network | |
CN111062973A (en) | Vehicle tracking method based on target feature sensitivity and deep learning | |
CN107945153A (en) | A kind of road surface crack detection method based on deep learning | |
CN106778687A (en) | Method for viewing points detecting based on local evaluation and global optimization | |
CN107368787A (en) | A kind of Traffic Sign Recognition algorithm that application is driven towards depth intelligence | |
CN103425986B (en) | Mammary gland tumor image characteristic extracting method based on edge neighborhood weighting | |
CN105938564A (en) | Rice disease recognition method based on principal component analysis and neural network and rice disease recognition system thereof | |
CN106408030A (en) | SAR image classification method based on middle lamella semantic attribute and convolution neural network | |
CN104598908A (en) | Method for recognizing diseases of crop leaves | |
CN106295124A (en) | Utilize the method that multiple image detecting technique comprehensively analyzes gene polyadenylation signal figure likelihood probability amount | |
CN102867195B (en) | Method for detecting and identifying a plurality of types of objects in remote sensing image | |
CN112069985B (en) | High-resolution field image rice spike detection and counting method based on deep learning | |
CN107563349A (en) | A kind of Population size estimation method based on VGGNet | |
CN104715244A (en) | Multi-viewing-angle face detection method based on skin color segmentation and machine learning | |
CN112699967B (en) | Remote airport target detection method based on improved deep neural network | |
CN104715251B (en) | A kind of well-marked target detection method based on histogram linear fit | |
CN104182985A (en) | Remote sensing image change detection method | |
CN103048329A (en) | Pavement crack detecting method based on active contour model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160323 |